I didn’t intend to comment at length on this subject just yet, but there is evidently some disagreement on the meaning of ‘group selection’, and it may help if I say what I mean by it myself. Sorry, it’s another long one!
A quick disclaimer: nothing in this note is concerned with group selection of cultural traits. It deals only with plain old-fashioned genes. For ‘cultural evolution’ see my earlier note.
The Problem
Many animals, including man, behave in ways that appear to reduce their chances of survival and reproduction. For example, they give alarm calls when they see a predator, even though this may lead the predator towards them. Or they give food to the offspring of another animal. Or they fail to kill a mating rival when they have an opportunity to do so. And so on.
Some of this behaviour may simply be accidental, an imperfection of nature.
However, some ‘altruistic’ actions seem to be part of the evolved behavioural repertoire of the species concerned. This is a problem, because at first sight it seems that natural selection can only favour the survival and reproduction of an individual organism and its direct descendants.
Historically, biologists in the generations after Darwin attempted to resolve the problem by arguing that among social animals altruistic behaviour would be beneficial to the group in which they lived, and that this group benefit might offset the individual disadvantage. Darwin himself said a few words along these lines. But nobody worked through the details of the process. Then in the early 1960s a British biologist called Wynne-Edwards wrote a long book arguing that many common animal behaviours, such as territoriality, could be explained by group advantage. Wynne-Edwards did not use much mathematics, but he did at least set out the details of the process explicitly for the first time. The problem was that the process he described was absurd, and the biological community turned strongly against group selection (see G. C. Williams, ‘Adaptation and Natural Selection’, 1964).
This left altruistic behaviour to be explained in some other way. Often the paradox can be resolved by showing that the behaviour is not altruistic at all. For example, why kill a defeated rival if in doing so you risk injury? More subtly, altruism might be ‘reciprocal’ and mutually beneficial, if individuals interact frequently enough to exchange benefits. But not every example can be dismissed as ‘disguised selfishness’. The problem is how to explain behaviour which involves a net fitness cost C to the performer, while conferring a net benefit B (in aggregate greater than C), on one or more other individuals. On the face of it, any gene which promotes such behaviour, without conferring any other benefit on its possessor, should be eliminated by natural selection. If all the other individuals in the population possessed the same gene, there would be no immediate problem, because they would collectively receive the greater benefit B. But there is still the problem of explaining how such a gene could have spread in the first place. And even if it did, one would expect selfish variants to emerge, which would receive the benefit without paying the cost.
We must therefore assume that there is a mixed population of ‘altruists’ and
‘non-altruists’. If altruists confer their benefit on other members of the species at random, altruism will be eliminated. The benefit B will be divided, in the long run, between altruists and non-altruists in proportion to their numbers in the population (or to be precise, the population excluding the particular altruist concerned), while the cost C falls only on the altruists. To enable altruism to survive, there must be some way of concentrating benefits disproportionately on those individuals who carry the genes for the altruistic behaviour.
Selective Behaviour
This could occur by means of selective behaviour, if the altruist discriminated in favour of other altruists. The most obvious way in which this might happen would be if the altruist gave the benefit preferentially to his own genetic relatives, who would be more likely than non-relatives to carry the relevant
genes. Alternatively, the altruist might give the benefit to other altruists who are not relatives, if he observes them performing altruistic behaviour, or because the genes responsible for altruism also produce some other observable feature (Dawkins’s ‘Green Beard’ effect). For the reasons given by Dawkins, only discrimination in favour of relatives is at all likely.
Group Structure
But even without discriminatory behaviour, benefits might be conferred disproportionately on other altruists if for any reason altruists tend to be
concentrated together. This is where group structure comes in. If the species is divided into groups, some groups may contain more altruists than others, and in these groups altruism may be rewarded. However, it is not sufficient merely for groups to be formed at random. In the long run, this would just be another way for altruists to distribute their benefits randomly to other members of the population. (See the explanation in Hamilton, ‘Narrow Roads’, vol. 1, p.334-5 – and good luck with the algebra.)
Nevertheless, ‘chance’ would sometimes produce a concentration of altruists
greater than the ‘long run expectation’, and this might give altruism a foothold to get going. (Maynard Smith’s 1964 ‘Haystack’ model seems to be of this kind, though he does not express it in these terms.) Calculations and computer simulations of various models show that this process can work, but it requires a fairly narrow range of conditions to do so, and current enthusiasts for group selection do not seem to rely on it.
If not chance, what? The most obvious answer is that, in many species, genetic relatives are likely to live near each other. They do not even have to be close relatives in the usual sense – as Hamilton pointed out, in any group, if it is isolated from immigration, genetic relatedness will build up by inbreeding. (It is an empirical matter whether most groups are isolated enough for this to happen.) Hamilton also pointed out that there could be reasons other than genetic relatedness for altruists to associate together – they might recognise fellow altruists as such, or they might settle together because the genes for altruism have some other effect on habitat preference. Sober and Wilson, in ‘Unto Others’, have stressed that altruists may congregate together because everybody, including altruists, wants to be near an altruist. This is a good point, and in a spirit of friendliness (altruism?) towards group-selectionists, I offer the following simple model, which so far as I know is new. Animals live in squares on a chessboard, with at most one animal to each square (some squares are empty). Each animal may therefore have up to eight immediate neighbours. The animals follow two simple rules: if they have three or more altruists as immediate neighbours, they stay where they are; if they do not, they move to a randomly chosen empty square. It is easy to see that some patterns of settlement will be stable under these rules – e.g. a block of four or more altruists in a square or rectangle – while others are not. Altruists will tend to clump together far more than would be expected by ‘chance’. This is of course a fantasy. I doubt if animals often behave like this. Perhaps more important, if altruists tend to associate together they will also tend to interbreed (assuming sexual reproduction), so the model quickly collapses into one of genetic relatedness again.
Synergy
Finally, a quite different way of concentrating benefits on fellow altruists would be if the benefits of altruism increase disproportionately in a ‘synergistic’ way when the number of altruists interacting is high (Maynard Smith, ‘Evolutionary Genetics’). It is surprising that the ‘groupies’ have not made more of this possibility. Personally, I think it could be important in human evolution.
Semantics
The odd thing about this whole debate is that most biologists agree pretty much on what is theoretically possible, and even on what is most likely to occur in practice. Overwhelmingly the most likely way for altruism to be favoured by selection is for genetic relatives to confer benefits on each other, either preferentially, or as a by-product of geographical concentration.
A large part of the debate is therefore purely semantic. Should a benefit conferred on relatives be described as ‘group selection’? Sober and Wilson and their own ‘groupies’ are probably the only biologists who think it should. There is of course no absolute right and wrong about a semantic issue, but historically this process is not what was originally meant by group selection, and in long-established usage it is known as ‘kin selection’. (Hamilton, in a conciliatory spirit, once suggested calling it ‘kin group selection’.) What decides the issue for me is that the term ‘group selection’ has a proven capacity for causing confusion and woolly thinking, and it should be avoided if at all possible. I would certainly confine it to cases involving (mainly) non-relatives, but even then the various models are so different that I do not think a rag-bag term like ‘group selection’ serves any useful purpose.
DAVID BURBRIDGE
Comments are closed.